Stanford professor warns: models are increasingly opaque and brittle, hasty deployment amplifies misalignment risk

anshulkundaje · x · 2026-09-16

Stanford professor Anshul Kundaje argues the current AI paradigm is brittle and heavily dependent on training data choices, with model behavior unpredictable in unintentionally misaligned zones. As models grow more powerful and opaque, he warns these risks can't be brushed off—especially with humans deploying them hastily in flawed environments with poor safeguards.

Related event: Stanford Scholars Warn Stronger Models Are More Opaque and Fragile(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →